SPCE061A is a 16-bit single-chip designed by Sunplus. The MCU is rich in resources and has a very high cost performance. The MCU has built-in 2-way DA conversion, 8-channel AD conversion and online simulation, and has 16×16-bit multiplication. The DSP function of the operation and inner product operation, which facilitates the compression encoding and decoding of complex speech digital signals, and can also be used as a digital filter. These features facilitate our application in the field of speech processing, especially speech recognition.
This article refers to the address: http://
Speech recognition technology has broad application prospects at all levels. Computer software fields, such as: voice commands, voice input, dialogue systems, inquiry systems, teaching software, game software, etc.; consumer electronics products, such as: electronic notebooks, voice-activated toys, voice dialing functions, etc.; industrial products Fields, such as: mobile phones for vehicles, navigation systems for vehicles, etc.; in the field of telephone systems, such as voice recognition switchboard services, voice dialing, voice booking reservations, etc.
1 Basic principles of speech recognition
Speech recognition is to let the machine understand what people are saying. The basic framework is shown in Figure 1 and Figure 2. Figure 1 is a portion of a speech training model that produces a standard speech reference sample from known speech signals via End Point Detection and Feature Extraction.
Figure 2 is a portion of speech recognition. The speech signal to be tested is obtained by the same processing steps as in FIG. 1, and then compared with the aforementioned standard speech reference sample to find the most similar reference sample as the identification result.
2 speech recognition classification
According to the restrictions of the user, it can be divided into specific person speech and non-specific person speech recognition.
2.1 Specific person speech recognition
Before using a specific person's speech recognition system, the user's speech reference sample must first be stored in a comparison database, that is, the specific person speech recognition system must first perform the training learning step of FIG. 1 before use.
2.2 Non-specific person speech recognition
You don't need to learn before you use this system, you can use it directly. A set of best speech recognition systems can perform speech recognition without learning, but usually the recognition rate is relatively low.
In addition, whether the continuity of the speech mode can be further divided into discontinuous speech recognition and continuous speech recognition. For non-continuous speech, each word recognized must be recognized separately; while continuous speech recognition can be humanized speech recognition in a generally fluent way of speaking, but it is difficult to achieve good due to the problem of connected sound. The recognition effect.
3 SPCE061A chip features
SPCE061A is a 16-bit MCU based on μ'nSP core. Its chip characteristics are as follows: (1) Operating voltage: The core operating voltage VDD is 3.0~3.6V (CPU), and the I/O port working voltage VDDH is 2.4~5.5V. (I/O); (2) CPU clock: 0.32MHz ~ 49.152MHz; (3) built-in 2K word SRAM and 32K flash ROM; (4) the system is in standby state (clock is stopped), power consumption is less than 2μ; (5) With touch-key wake-up function; (6) 32-bit general-purpose programmable input/output port; (7) Two 16-bit programmable timers/counters (automatically preset initial count value); (8) 7 Channel 10-bit voltage analog-to-digital converter (ADC) and single-channel acoustic analog-to-digital converter; (9) 2 10-bit DAC (digital-to-analog conversion) output channels; (10) 14 interrupt sources from timers A/B, time base, 2 external clock source inputs, key wake-up; (11) with serial device interface; (12) low voltage reset (LVR) function and low voltage monitoring (LVD) function; (13) built-in online Simulation (ICE, In-Circuit Emulator) interface.
The internal structure block diagram of SPCE061A is shown in Figure 3.
4 SPCE061A speech recognition application
4.1 Hardware Circuit Brief
The hardware implementation of SPCE061A's D/A conversion channel is roughly two types, one is to directly use digital-to-analog converter (DAC), and the other is to directly use pulse width modulation (PWM) driving, SPCE061A audio output structure. It consists of two DAC channels or one PWM drive channel. Both types of implementations are essentially digital-to-analog conversions, which convert digital signals into current analog signal outputs. The audio output of SPCE061A adopts dual-channel digital-to-analog conversion DAC mode, that is, the digital signal is converted into a 3mA-driven current analog signal output through a 10-bit DAC. The voice prompt output is directly used by the DAC channel, and after being amplified by the signal, it is output by the speaker.
The speech recognition circuit is relatively simple, and the MIC uses an electret microphone. The electret microphone has the advantages of simple structure, light weight, small size, non-directionality, wide frequency response and good fidelity. Its bias voltage is provided by the VMIC pin of SPCE061A.
4.2 System Software
Speech recognition is divided into two processes: speech model training and speech recognition. We refer to the standard mode storage space as the "thesaurus" and the standard mode as the "model". The so-called speech model training, the spectrum to be identified is to be analyzed, and the feature parameters are extracted as the standard mode of recognition. The process of speech recognition is to extract the feature parameters of the voice command, compare it with the template in the thesaurus, and take the sample command number with the best similarity as the recognition result. Lingyang Technology Co., Ltd. provides voice model training tools and speech recognition library functions. Each time, 30 voice commands can be recognized. More commands can be divided into multiple groups. The voice recognition process is shown in Figure 4.
Among them, the BSR-GetResult function is the startup identification. When no command is recognized, it returns 0; the recognizer stops, uninitializes or recognizes that it is not activated and returns -1; when the recognition fails, it returns -2; when it is recognized, it returns the command. Serial number. Part of the program source code is as follows.
The above program code is an application of SPCE061A in an electric trolley based on a specific human intelligent speech recognition. A large number of experimental tests have shown that SPCE061A is applied to the field of speech recognition, and the system performance is stable, sound quality and recognition effect are better. However, when the ambient noise or interference signal is higher than the speech signal, the system will not be able to perform correct speech recognition, and the background noise processing and its engineering will actually be further improved.
The Rockwell Controllogix processor provides an optional user Memory Module (750K to 8M bytes) that can solve application problems with a large number of input and output points systems (up to 4000 analog and 128000 digital bits). The processor can control local input and output and remote input and output. The processor can monitor input and output in the system via Ethernet EtherNet / IP, ControlNet ControlNet, DeviceNet DeviceNet, and Remote I/O Universal Remote I / O.
When there are multiple processor modules in the Controllogix chassis, and even if there are multiple processor modules in the ControlNet network of the control network, all processors can read input values from all input modules. Any one processor can also control any specific output module. The system configuration specifies which processor is controlled by each output module.
The ControlLogix system is a rack-mounted, modular installation. The Controllogix I/O Modules are modularly mounted. The power module is mounted directly to the left side of the ControlLogix chassis. The Controllogix chassis is available in five types of 4, 7, 10, 13 or 17 slots. The module can be inserted in any slot of the rack. The maximum number of channels for Controllogix I/O modules is 32 channels. The mechanical lock of the removable terminal block of each module prevents the application of erroneous voltages to the module. Input and output modules can be hot plugged.
Rockwell Allen-Bradley: SLC500/1747/1746 MicroLogix/1761/1763/1762/1766/1764
CompactLogix/1769/1768
Logix5000/1756/1789/1794/1760/1788,PLC-5/1771/1785 and so on.
Rockwell Allen-Bradley
Rockwell Allen-Bradley,Processor Controll,Rockwell Automation Allen-Bradley,Allen-Bradley Equipments
Xiamen The Anaswers Trade Co,.LTD , https://www.answersplc.com